How to integrate a deep neural network for spell checking in the RASA framework?

I have built a spell checker specific for my project using RNN. I am trying to integrate it with RASA.

I have explored a couple of options to do that:-

  1. User input spell checking before sending to RASA
  2. Using vectorized words from RASA to do spell checking

Any suggestions on this?

Welcome to the forum @Vikram_kamboj!

You could write your own custom component (see Custom NLU Components) and add it to your pipeline. Depending on how your component works, you can operate on tokens or any other part of the message.

1 Like